[Foundation-l] The average voter and voting systems (was Re: Notice of the results of the WMF Board of Trustees election)

[Daniel Arnold]

On Friday 13 July 2007 08:11:49 Michael Snow wrote:
> Considering that the election was for three seats on the board, using
> approval voting, the fact that the average voter voted for barely more
> than three candidates is quite disappointing. There are three
> possibilities to explain this:
> [...]
> Explanation #3 is a little harder to evaluate. I'm sure some people felt
> there was a shortage of candidates they felt comfortable voting for, and
> we could always wish for even better options to choose. However, even if
> this was a widespread sentiment, there's hardly any indication that
> voters agreed on which candidates. The vote differences between
> candidates are not terribly large at most positions. I would say the
> results don't reflect people seeing a clear gap between "qualified" and
> "unqualified".

I had exactly one candidate I was able to support without any doubt (however 
this candidate is not among the elected members ;-).

So I had two second places left as I wanted to express my maximum will (voting 
for all is almost the same as voting for none) and this took me personally a 
hard time to take a decission based on facts and not on (e.g. nationality) 
taste.

> Instead, I would suggest keeping our present voting system in some form,
> but also making the decision faced by the voter more explicit. This can
> be done by presenting the voter with two choices for each candidate -
> Yes and No, or Approve and Disapprove if you prefer. This makes clear
> what you want the voter to tell you, and that the voter is answering
> this question independently for each candidate. [...]

This sounds like a good suggestion and sadly is the only really useful that is 
not so strongly based on personal tastes and beliefs. I was astonished by the 
mass of emails that wanted to promote their pet vote system just because they 
like it and not because there are independent mathematical proofs and 
usability studies that support their adoption.

There are experts that do mathematical analysis of vote systems. One famous 
mathematician in that subject is Michael Balinski from Paris, France.

http://ceco.polytechnique.fr/home/balinski/FR

Please read papers like the ones from him prior to any further decission on 
vote systems. If you hesitate to read through original research scientific 
papers there was a very good article in the German branch of Scientfic 
American called Spektrum der Wissenschaft: "Wer wird Präsident?", September 
2002, pp. 74-79 (French original: Le scrutin. Pour la science, 294, pp. 
46-51.) In this article he analyses vote systems for presidential elections 
(one winner elections). He was able to proof that every curent presidential 
vote system is not totally fair but that approval voting is the best one one 
which is both a usable and as fair as possible.

However this article did not cover the question whether approval voting is 
favourable for multiple winners where you have a ranking within your choice.

Another mathematician Friedrich Pukelsheim did analyse vote systems of 
parliamentary elections and did find a better one (better than 
d'Hondt, "winner take all" and what not) which is now being used in Zurich, 
Switzerland (Spektrum der Wissenschaft: "Die Mathematik der doppelten 
Gerechtigkeit", Volume 04/2007, p. 76 ff). You can find on his web site also 
a caculator for this system:

http://www.math.uni-augsburg.de/stochastik/pukelsheim/welcome.html (his web 
site)
http://www.math.uni-augsburg.de/stochastik/bazi/welcome.html (the Java 
caculator)

So maybe there were other experts that did find a good vote system for our use 
case. But please let us stop telling each other our personal preferences 
without any experts backing.

Cheers, Arnomane
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.wikimedia.org/pipermail/wikimedia-l/attachments/20070715/b2e49f97/attachment-0001.pgp>